Most hard support questions are about how the product works, and the answer is in the code. Claude Code can read the code, but it only works while you are at the keyboard and starts each session fresh. Afterchat is the part that stays: the chat on your site that answers around the clock, the inbox, and the knowledge vault that keeps every answer. Connected over MCP, Claude Code works the inbox with your repo open.
Set it up
-
Add the server. In a terminal:
claude mcp add --transport http --scope user afterchat https://afterchat.chat/mcp--scope usermakes it available in every project. Leave it out to add it to the current project only. Sign in. Restart Claude Code, run
/mcp, pick afterchat and choose Authenticate. A browser opens on Afterchat. Signing in makes your account if you don't have one, and you pick the workspace and untick anything Claude should not do.Check it works. Ask Claude to "check Afterchat status". It tells you the workspace, what it may do and what is waiting.
Or paste this into Claude Code and let it do the steps itself:
Connect Afterchat to this agent: fetch https://afterchat.chat/agent-setup.md and follow it.
Starting with an empty workspace? Run /mcp__afterchat__vault in your repo first, and Claude writes the vault from your code and docs for you to review.
A morning in the inbox
Open Claude Code in your repo and run /mcp__afterchat__inbox, or just ask it to check the Afterchat inbox. It lists the conversations waiting on you, technical ones first. For each one it reads the conversation, searches the vault, opens the code it needs and drafts a reply into the inbox.
The reply waits there for your OK, with a line saying what Claude checked and what it saved. You approve it in Afterchat, or edit it first. When the conversation closes, what Claude learned is kept as an entry that points at the files it read, and the chat on your site answers that question by itself from then on.
The commands
| Command | What it does |
|---|---|
/mcp__afterchat__inbox | Works through the conversations waiting on you, technical ones first |
/mcp__afterchat__investigate | Digs into one conversation with the code and your own tools |
/mcp__afterchat__gaps | Answers the questions the vault is missing, from the code |
/mcp__afterchat__vault | Builds or improves the vault: topics, entries, merges and guides |
/mcp__afterchat__review | Checks unchecked answers and tidy-up suggestions against the code |
/mcp__afterchat__triage | Lays out the ranked requests, for deciding what to build |
/mcp__afterchat__build | Loads a request's brief to build from |
/mcp__afterchat__ship | Tells everyone who asked and updates the vault, once you approve |
What waits for you
Claude works as you, with what you allowed when you signed in. Replies to customers wait as drafts unless an owner lets them go out at once, and hold rules keep prices, limits, refunds, legal terms and anything quoting your code with a person whatever the setting. Public changes and shipping a request wait on the Coding agent page, and every call is in the log with an undo. The full list is on the MCP server page.
On our own support
We run FanWake's support this way. Since 26 October 2026 Claude Code has made over 800 calls to it, reading chats, drafting replies and keeping the vault. On 31 October one session rewrote the vault from FanWake's code and docs, 152 entries in 10 topics, which were reviewed and approved as one batch. The whole story, with the numbers, is in the write-up on the blog.